I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

---
View in Courier:



+-------------------3.6V to MOSFETS
|
+----[1N5817]--+---3.3V to PIC
| |+
[BAT] [BFC]
| |C1
+---------------+
|
GND

The 1N5817 is a Schottky diode and will drop about 0.3V.
The BFC is there to keep the PIC supply from dropping too far and
resetting the PIC when the MOSFETS supply power to the load.

The value of the capacitance is:

IdT
C = -----
dv

Where C is the capacitance in Farads,

I is the current in Amperes,

dT is the length of time you require the capacitor to supply
current to the PIC and,

dV is the PIC supply voltage minus the lowest voltage the PIC
supply can fall to before the PIC resets.

So, making some assumptions:


IdT 5E-2A * 5E-6µs
C1 = ----- = ---------------- = 2.5E-6F = 2.5µF
dv 3.3V - 3.2V

Which means that if the battery voltage falls below 3.6V during the
load transient, a 2.5µF cap will keep the PIC supply above 3.2V for
5µs or longer if the PIC circuitry draws 50mA or less. Of course
you could use a larger cap, say 10µF or so and run farther away from
the hairy edge.


--
JF

"John Fields" wrote in message
...
On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

---
View in Courier:



+-------------------3.6V to MOSFETS
|
+----[1N5817]--+---3.3V to PIC
| |+
[BAT] [BFC]
| |C1
+---------------+
|
GND

The 1N5817 is a Schottky diode and will drop about 0.3V.
The BFC is there to keep the PIC supply from dropping too far and
resetting the PIC when the MOSFETS supply power to the load.

The value of the capacitance is:

IdT
C = -----
dv

Where C is the capacitance in Farads,

I is the current in Amperes,

dT is the length of time you require the capacitor to supply
current to the PIC and,

dV is the PIC supply voltage minus the lowest voltage the PIC
supply can fall to before the PIC resets.

So, making some assumptions:


IdT 5E-2A * 5E-6µs
C1 = ----- = ---------------- = 2.5E-6F = 2.5µF
dv 3.3V - 3.2V

Which means that if the battery voltage falls below 3.6V during the
load transient, a 2.5µF cap will keep the PIC supply above 3.2V for
5µs or longer if the PIC circuitry draws 50mA or less. Of course
you could use a larger cap, say 10µF or so and run farther away from
the hairy edge.


--
JF

I like it! For the surface mount application, though, I'll use a BAT54A (or
BAT54C) for the Schottky diode and probably bump the capacitance to about
10uF to 22uF. The BAT54A (series) can take up to about 200mA without
popping. SOT23-3 format, too.

Thanks! I'll try that out.

Dave

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.



C30 is not a good idea.

John

John Larkin wrote:

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Normally, I'm not a pest about spelling, but .......

C30 is not a good idea.

John

Do you think C30 is causing the sag? I think that might slow down Q2,
but the sag is caused by having to charge C31. What kind of 3.6V battery
is that?

Hint to the OP: If you can, set your scope's trigger delay to -1 units
so we can see what things looked like before they fell off the edge of
the earth.

--
Paul Hovnanian
------------------------------------------------------------------
"Yee-Ha!" is not an adequate foreign policy.

On Mon, 11 Jun 2007 22:31:43 -0700, "Paul Hovnanian P.E."
wrote:

John Larkin wrote:

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Normally, I'm not a pest about spelling, but .......

C30 is not a good idea.

John

Do you think C30 is causing the sag? I think that might slow down Q2,
but the sag is caused by having to charge C31. What kind of 3.6V battery
is that?

Hint to the OP: If you can, set your scope's trigger delay to -1 units
so we can see what things looked like before they fell off the edge of
the earth.

It could add "gain" to the sag, by increasing Rds-on when the source
voltage dips. Just get rid of it, or use a s-g cap if you want to slow
down turnon.

But, looking at the dip waveform, it's awfully fast for being a
"circuit" issue. Looks more like a layout problem. I wonder if this is
on a proto board.

John

On Mon, 11 Jun 2007 19:42:49 -0600, "starfire"
wrote:


"John Fields" wrote in message
.. .
On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

---
View in Courier:



+-------------------3.6V to MOSFETS
|
+----[1N5817]--+---3.3V to PIC
| |+
[BAT] [BFC]
| |C1
+---------------+
|
GND

The 1N5817 is a Schottky diode and will drop about 0.3V.
The BFC is there to keep the PIC supply from dropping too far and
resetting the PIC when the MOSFETS supply power to the load.

The value of the capacitance is:

IdT
C = -----
dv

Where C is the capacitance in Farads,

I is the current in Amperes,

dT is the length of time you require the capacitor to supply
current to the PIC and,

dV is the PIC supply voltage minus the lowest voltage the PIC
supply can fall to before the PIC resets.

So, making some assumptions:


IdT 5E-2A * 5E-6µs
C1 = ----- = ---------------- = 2.5E-6F = 2.5µF
dv 3.3V - 3.2V

Which means that if the battery voltage falls below 3.6V during the
load transient, a 2.5µF cap will keep the PIC supply above 3.2V for
5µs or longer if the PIC circuitry draws 50mA or less. Of course
you could use a larger cap, say 10µF or so and run farther away from
the hairy edge.


--
JF

I like it! For the surface mount application, though, I'll use a BAT54A (or
BAT54C) for the Schottky diode and probably bump the capacitance to about
10uF to 22uF. The BAT54A (series) can take up to about 200mA without
popping. SOT23-3 format, too.

---
Watch Vf with If.


--
JF

On Tue, 12 Jun 2007 08:12:10 -0700, John Larkin
wrote:

On Mon, 11 Jun 2007 22:31:43 -0700, "Paul Hovnanian P.E."
wrote:

John Larkin wrote:

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Normally, I'm not a pest about spelling, but .......

C30 is not a good idea.

John

Do you think C30 is causing the sag? I think that might slow down Q2,
but the sag is caused by having to charge C31. What kind of 3.6V battery
is that?

Hint to the OP: If you can, set your scope's trigger delay to -1 units
so we can see what things looked like before they fell off the edge of
the earth.

It could add "gain" to the sag, by increasing Rds-on when the source
voltage dips. Just get rid of it, or use a s-g cap if you want to slow
down turnon.

But, looking at the dip waveform, it's awfully fast for being a
"circuit" issue. Looks more like a layout problem. I wonder if this is
on a proto board.

---
The OP reports that it occurs when there's no load on the big MOSFET
and he sends a high to the gate of the first MOSFET, so it could be
due to the gate capacitance of the first MOSFET charging up.


--
JF

John Fields wrote:

On Tue, 12 Jun 2007 08:12:10 -0700, John Larkin
wrote:

On Mon, 11 Jun 2007 22:31:43 -0700, "Paul Hovnanian P.E."
wrote:

John Larkin wrote:

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Normally, I'm not a pest about spelling, but .......

C30 is not a good idea.

John

Do you think C30 is causing the sag? I think that might slow down Q2,
but the sag is caused by having to charge C31. What kind of 3.6V battery
is that?

Hint to the OP: If you can, set your scope's trigger delay to -1 units
so we can see what things looked like before they fell off the edge of
the earth.

It could add "gain" to the sag, by increasing Rds-on when the source
voltage dips. Just get rid of it, or use a s-g cap if you want to slow
down turnon.

But, looking at the dip waveform, it's awfully fast for being a
"circuit" issue. Looks more like a layout problem. I wonder if this is
on a proto board.

---
The OP reports that it occurs when there's no load on the big MOSFET
and he sends a high to the gate of the first MOSFET, so it could be
due to the gate capacitance of the first MOSFET charging up.

It appears as though the ringing is due either to layout, or the loss of
drive to the Q2 gate when the voltage collapses. The collapse itself
appears to be due to half of C10's charge getting dumped into C31. The
recovery time constant looks like a battery with an ohm or two internal
resrstance recharging it once Q2 shuts off.

There's no way to keep this circuit from chopping the 3.6 V source in
half without a hefty (low impedance) source. Fields' BFC solution might
be the only thing to keep the uP alive during the transient, but the
drive power for Q2's drive might have to come from there as well to keep
it on during the transient.

--
Paul Hovnanian
------------------------------------------------------------------
Plaese porrf raed befre postng.

On Tue, 12 Jun 2007 11:06:04 -0500, John Fields
wrote:

On Tue, 12 Jun 2007 08:12:10 -0700, John Larkin
wrote:

On Mon, 11 Jun 2007 22:31:43 -0700, "Paul Hovnanian P.E."
wrote:

John Larkin wrote:

On Mon, 11 Jun 2007 17:57:48 -0600, "starfire"
wrote:

I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Normally, I'm not a pest about spelling, but .......

C30 is not a good idea.

John

Do you think C30 is causing the sag? I think that might slow down Q2,
but the sag is caused by having to charge C31. What kind of 3.6V battery
is that?

Hint to the OP: If you can, set your scope's trigger delay to -1 units
so we can see what things looked like before they fell off the edge of
the earth.

It could add "gain" to the sag, by increasing Rds-on when the source
voltage dips. Just get rid of it, or use a s-g cap if you want to slow
down turnon.

But, looking at the dip waveform, it's awfully fast for being a
"circuit" issue. Looks more like a layout problem. I wonder if this is
on a proto board.

---
The OP reports that it occurs when there's no load on the big MOSFET
and he sends a high to the gate of the first MOSFET, so it could be
due to the gate capacitance of the first MOSFET charging up.



Maybe so. Then a series gate resistor, and one in the drain too,
maybe, would reduce the spike current.

John

"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Values for C9 and C10 are too small (common error). C9 should be at least
10 uF
Try 47 to 100 uF for C10

gb

"w9gb" wrote in message
. ..
"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.

Values for C9 and C10 are too small (common error).

My error -- Input capacitor (C10) should be at least 10 uF

Output capacitor (C9) -- Try 47 to 100 uF

"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.




Apologies for not responding with fix solution results but I've been pulled
away on other things for the last couple of days.

I will try many of the fix solutions tomorrow night and post results.

Thanks everyone.

Dave

"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.


I'm posting some traces from some of the implemented suggested
modifications:

trace 1: no load from switched 3.6VDC to ground
trace 2: 1K-ohm load from switched 3.6VDC to ground
trace 3: 100-ohm load from switched 3.6VDC to ground
trace 5: 10-ohm load from switched 3.6VDC to ground

I installed 10uF caps across the exisitng input 1uF caps on the input and
output of the TC1108 voltage regulator.
I also installed a 10uF cap across the existing 0.1uF cap on the Vcc and
ground leads of the PIC right at the power pins.
I also removed C30 (the 0.1uF cap on the drain of the N-Channel MOSFET and
the gate of the P-Channel MOSFET.

As can be seen in the traces, the dip seems to be unrelated to the amount of
switched current. The dip is also reduced significantly in amplitude now.
The switching action works now every time I tried it. This was all very
welcome news!

I used 10uF caps because that's what I had on hand and they were small
enough to fit nicely across the pins of the voltage regulator. I will try
to find some 22uF (or so) caps to see if I can reduce the dip even further.
This is very hopeful...

I tried the series diode with a BFC from the PIC Vcc to ground on a separate
breadboard to convince myself this will work, also. It looks like this
approach would work, too. I'm a little reluctant in going that way if I
need to get analog inputs from an accelerometer (which is a proposed add-on
to this system). If I use a Schottky diode drop from the raw battery to the
PIC (along with a BFC to ground), the PIC Vcc could vary significantly
dependant on the current load going into the PIC and the charge status of
the battery. I will keep this one in the back pocket, though...

Thanks immensely to all who responded.

Dave

On Thu, 14 Jun 2007 15:06:52 -0600, "starfire"
wrote:


"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from Paul
Hovnarian. Thanks for the suggestion, Paul.


I'm posting some traces from some of the implemented suggested
modifications:

trace 1: no load from switched 3.6VDC to ground
trace 2: 1K-ohm load from switched 3.6VDC to ground
trace 3: 100-ohm load from switched 3.6VDC to ground
trace 5: 10-ohm load from switched 3.6VDC to ground

I installed 10uF caps across the exisitng input 1uF caps on the input and
output of the TC1108 voltage regulator.
I also installed a 10uF cap across the existing 0.1uF cap on the Vcc and
ground leads of the PIC right at the power pins.
I also removed C30 (the 0.1uF cap on the drain of the N-Channel MOSFET and
the gate of the P-Channel MOSFET.

As can be seen in the traces, the dip seems to be unrelated to the amount of
switched current. The dip is also reduced significantly in amplitude now.
The switching action works now every time I tried it. This was all very
welcome news!

I used 10uF caps because that's what I had on hand and they were small
enough to fit nicely across the pins of the voltage regulator. I will try
to find some 22uF (or so) caps to see if I can reduce the dip even further.
This is very hopeful...

I tried the series diode with a BFC from the PIC Vcc to ground on a separate
breadboard to convince myself this will work, also. It looks like this
approach would work, too. I'm a little reluctant in going that way if I
need to get analog inputs from an accelerometer (which is a proposed add-on
to this system). If I use a Schottky diode drop from the raw battery to the
PIC (along with a BFC to ground), the PIC Vcc could vary significantly
dependant on the current load going into the PIC and the charge status of
the battery. I will keep this one in the back pocket, though...

---
Since the ADC in the PIC is ratiometric, (I believe) there will be
no errors due to changes in Vcc if you excite the accelerometer with
Vcc.


--
JF

"John Fields" wrote in message
...
On Thu, 14 Jun 2007 15:06:52 -0600, "starfire"
wrote:


"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from
Paul
Hovnarian. Thanks for the suggestion, Paul.


I'm posting some traces from some of the implemented suggested
modifications:

trace 1: no load from switched 3.6VDC to ground
trace 2: 1K-ohm load from switched 3.6VDC to ground
trace 3: 100-ohm load from switched 3.6VDC to ground
trace 5: 10-ohm load from switched 3.6VDC to ground

I installed 10uF caps across the exisitng input 1uF caps on the input and
output of the TC1108 voltage regulator.
I also installed a 10uF cap across the existing 0.1uF cap on the Vcc and
ground leads of the PIC right at the power pins.
I also removed C30 (the 0.1uF cap on the drain of the N-Channel MOSFET and
the gate of the P-Channel MOSFET.

As can be seen in the traces, the dip seems to be unrelated to the amount
of
switched current. The dip is also reduced significantly in amplitude now.
The switching action works now every time I tried it. This was all very
welcome news!

I used 10uF caps because that's what I had on hand and they were small
enough to fit nicely across the pins of the voltage regulator. I will try
to find some 22uF (or so) caps to see if I can reduce the dip even
further.
This is very hopeful...

I tried the series diode with a BFC from the PIC Vcc to ground on a
separate
breadboard to convince myself this will work, also. It looks like this
approach would work, too. I'm a little reluctant in going that way if I
need to get analog inputs from an accelerometer (which is a proposed
add-on
to this system). If I use a Schottky diode drop from the raw battery to
the
PIC (along with a BFC to ground), the PIC Vcc could vary significantly
dependant on the current load going into the PIC and the charge status of
the battery. I will keep this one in the back pocket, though...

---
Since the ADC in the PIC is ratiometric, (I believe) there will be
no errors due to changes in Vcc if you excite the accelerometer with
Vcc.


--
JF

Good point. I will try an experiment with the breadboard system and look at
that. In addition to the accelerometer inputs, I have an op-amp configured
as a voltage follower monitoring the center tap between a pair of 100K 1%
resistors in series. One end of the string goes to the raw 3.6VDC input.
The other end is tied to ground. I will see how this voltage changes with
varying raw input voltage.

Dave

On Thu, 14 Jun 2007 18:08:08 -0600, "starfire"
wrote:


"John Fields" wrote in message
.. .
On Thu, 14 Jun 2007 15:06:52 -0600, "starfire"
wrote:


"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from
Paul
Hovnarian. Thanks for the suggestion, Paul.


I'm posting some traces from some of the implemented suggested
modifications:

trace 1: no load from switched 3.6VDC to ground
trace 2: 1K-ohm load from switched 3.6VDC to ground
trace 3: 100-ohm load from switched 3.6VDC to ground
trace 5: 10-ohm load from switched 3.6VDC to ground

I installed 10uF caps across the exisitng input 1uF caps on the input and
output of the TC1108 voltage regulator.
I also installed a 10uF cap across the existing 0.1uF cap on the Vcc and
ground leads of the PIC right at the power pins.
I also removed C30 (the 0.1uF cap on the drain of the N-Channel MOSFET and
the gate of the P-Channel MOSFET.

As can be seen in the traces, the dip seems to be unrelated to the amount
of
switched current. The dip is also reduced significantly in amplitude now.
The switching action works now every time I tried it. This was all very
welcome news!

I used 10uF caps because that's what I had on hand and they were small
enough to fit nicely across the pins of the voltage regulator. I will try
to find some 22uF (or so) caps to see if I can reduce the dip even
further.
This is very hopeful...

I tried the series diode with a BFC from the PIC Vcc to ground on a
separate
breadboard to convince myself this will work, also. It looks like this
approach would work, too. I'm a little reluctant in going that way if I
need to get analog inputs from an accelerometer (which is a proposed
add-on
to this system). If I use a Schottky diode drop from the raw battery to
the
PIC (along with a BFC to ground), the PIC Vcc could vary significantly
dependant on the current load going into the PIC and the charge status of
the battery. I will keep this one in the back pocket, though...

---
Since the ADC in the PIC is ratiometric, (I believe) there will be
no errors due to changes in Vcc if you excite the accelerometer with
Vcc.


--
JF

Good point. I will try an experiment with the breadboard system and look at
that. In addition to the accelerometer inputs, I have an op-amp configured
as a voltage follower monitoring the center tap between a pair of 100K 1%
resistors in series. One end of the string goes to the raw 3.6VDC input.
The other end is tied to ground. I will see how this voltage changes with
varying raw input voltage.

---
It should always be half of the raw input voltage, +/- 1% worst
case, since for:


.. E1
.. |
.. [R1|
.. |
.. +---E2
.. |
.. [R2]
.. |
.. GND



E1 * R2
E2 = ---------
R1 + R2



--
JF

"John Fields" wrote in message
...
On Thu, 14 Jun 2007 18:08:08 -0600, "starfire"
wrote:


"John Fields" wrote in message
. ..
On Thu, 14 Jun 2007 15:06:52 -0600, "starfire"
wrote:


"starfire" wrote in message
...
I'm posting the associated circuit and traces here, per request from
Paul
Hovnarian. Thanks for the suggestion, Paul.


I'm posting some traces from some of the implemented suggested
modifications:

trace 1: no load from switched 3.6VDC to ground
trace 2: 1K-ohm load from switched 3.6VDC to ground
trace 3: 100-ohm load from switched 3.6VDC to ground
trace 5: 10-ohm load from switched 3.6VDC to ground

I installed 10uF caps across the exisitng input 1uF caps on the input
and
output of the TC1108 voltage regulator.
I also installed a 10uF cap across the existing 0.1uF cap on the Vcc and
ground leads of the PIC right at the power pins.
I also removed C30 (the 0.1uF cap on the drain of the N-Channel MOSFET
and
the gate of the P-Channel MOSFET.

As can be seen in the traces, the dip seems to be unrelated to the
amount
of
switched current. The dip is also reduced significantly in amplitude
now.
The switching action works now every time I tried it. This was all very
welcome news!

I used 10uF caps because that's what I had on hand and they were small
enough to fit nicely across the pins of the voltage regulator. I will
try
to find some 22uF (or so) caps to see if I can reduce the dip even
further.
This is very hopeful...

I tried the series diode with a BFC from the PIC Vcc to ground on a
separate
breadboard to convince myself this will work, also. It looks like this
approach would work, too. I'm a little reluctant in going that way if I
need to get analog inputs from an accelerometer (which is a proposed
add-on
to this system). If I use a Schottky diode drop from the raw battery to
the
PIC (along with a BFC to ground), the PIC Vcc could vary significantly
dependant on the current load going into the PIC and the charge status
of
the battery. I will keep this one in the back pocket, though...

---
Since the ADC in the PIC is ratiometric, (I believe) there will be
no errors due to changes in Vcc if you excite the accelerometer with
Vcc.


--
JF

Good point. I will try an experiment with the breadboard system and look
at
that. In addition to the accelerometer inputs, I have an op-amp
configured
as a voltage follower monitoring the center tap between a pair of 100K 1%
resistors in series. One end of the string goes to the raw 3.6VDC input.
The other end is tied to ground. I will see how this voltage changes with
varying raw input voltage.

---
It should always be half of the raw input voltage, +/- 1% worst
case, since for:


. E1
. |
. [R1|
. |
. +---E2
. |
. [R2]
. |
. GND



E1 * R2
E2 = ---------
R1 + R2



--
JF

Thanks, John.

I was thinking more of how the PIC's ADC would convert (what code was
produced) with the varying input voltage and the series Schottky in line
with the Vcc of the PIC. I would think it should be fairly stable
independent of raw voltage since both the voltage divider string and the PIC
are riding the raw changes. As long as the current draw from the PIC
doesn't vary much (due to loads switching on or off causing more current to
be drawn into the Vcc pin). This could potentially cause a variation in the
voltage drop across the Schottky diode. That could shift the code for the
center point. At least that's the theory...

Dave


Dave